THINC-scaling scheme that unifies VOF and level set methods

Bin Xie; Feng Xiao; Lidong Cheng; Ronit Kumar

REVIEW 4 major objections 6 minor 25 references

THINC-scaling scheme that unifies VOF and level set methods

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The volume-of-fluid and level set interface methods are two views of one object: THINC-scaling converts each into the other, so one procedure keeps volume conservation and geometrical faithfulness at the same time.

desk verdict A clean, genuinely useful unification of VOF and level set via THINC scaling, but the paper undersells its own novelty and oversells geometric faithfulness without measuring it. read the letter →

arxiv 1908.04529 v1 pith:Q2GXA773 submitted 2019-08-13 math.NA cs.NAphysics.flu-dyn

Ronit Kumar , Lidong Cheng , Bin Xie , Feng Xiao This is my paper · ORCID

classification math.NAcs.NAphysics.flu-dyn MSC 65M0876M1276T99

keywords THINCscalingvolumeoffluidlevelsetmethodinterfacecapturingsemi-Lagrangianupdateconservationhigh-orderpolynomialunstructuredgrid

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that volume-of-fluid (VOF) and level set methods, usually built as completely separate solvers, are really two readings of the same field under a hyperbolic-tangent scaling. In the THINC-scaling scheme, a level set polynomial is passed through a tanh function whose cell average is forced to equal the given VOF value, and the level set value is recovered from that same tanh function by the inverse transformation. The VOF field is advanced by a conservative finite volume update, while the level set field is advanced by a semi-Lagrangian update from the reconstructed tanh function without extra spatial interpolation. On solid-body rotation and vortex-deformation benchmarks, on both Cartesian and triangular grids, the scheme reconstructs interfaces including filaments thinner than a cell, with L1 errors comparable to geometrical VOF methods. If the scheme works as claimed, multiphase flow codes would no longer need to couple two different interface representations and choose between conserving volume and resolving geometry.

What carries the argument

The machinery is the THINC scaling identity and its inverse: $H(x)=\frac{1}{2}(1+\tanh(\beta P(x)))$ with $P$ a polynomial matching the level set derivatives, and $\varphi(x)=\frac{1}{\beta}\tanh^{-1}(2H-1)$. The scheme's working object is the cell-wise THINC function constructed under the exact VOF constraint, whose unknown shift $\varphi_\Delta$ is fixed by solving for the cell average, making the piecewise zero contour of $P+\varphi_\Delta$ the sub-cell interface. This one function does double duty: the finite volume update of the cell-averaged VOF uses it directly, and the semi-Lagrangian level set update evaluates it at the departure point through the inverse scaling, so no separate reconstruction operator is introduced. The steepness parameter $\beta$ controls interface thickness, and the paper uses $\beta\approx 6$ for a three-cell-thick transition layer.

What would settle it

Run the Rider\u2013Kothe single-vortex test on a 128 by 128 grid for the full forward\u2013reverse cycle, measuring total VOF and the L1 error of the returned circle against the exact initial shape: if volume drifts beyond the reported L1 level or the tail region fails to return to the circle, the claimed simultaneous conservation and geometry are not achieved. A sharper check is to compare the stretched tail at $t=T/2$ on the same grid against a high-resolution reference; if the under-resolved filament is not reproduced, the polynomial reconstruction assumption fails.

Watch

Extended reading notes

Core claim

The central claim is that the THINC function $H(x)=\frac{1}{2}(1+\tanh(\beta P(x)))$, built from a level set polynomial $P$ constrained by the derivatives of the level set field, is exactly the object that connects the two methods. Given a cell-wise VOF value, the unknown shift $\varphi_\Delta$ in the polynomial argument is solved so that the cell average of $H$ matches the VOF value, which enforces volume conservation by construction. The inverse formula $\varphi(x)=\frac{1}{\beta}\tanh^{-1}(2H-1)$ then supplies the level set value at any point in the cell, including the departure point of a semi-Lagrangian trajectory, so the level set update needs no polynomial interpolation of the level set field itself. The zero contour of the shifted polynomial, called the Polynomial Surface of the Interface, gives the sub-cell interface geometry including position, normal, and curvature inside each interface cell. The paper argues that this two-way scaling unifies VOF and level set into a single conservative and geometrically faithful framework, and supports the argument with advection tests on structured and unstructured grids.

Load-bearing premise

The load-bearing premise is that the cell-wise polynomial fit to the discrete level set values remains accurate inside the cell even when an interface feature is thinner than the grid; if the fit is poor, the inverse THINC step hands an incorrect level set value to the next cycle, and the reinitialization may not fully repair it.

Editorial extensions

If this is right

A single implementation can carry both the VOF field and the level set field, eliminating the need for two separate advection solvers and specialized coupling logic.
The sub-cell interface from the polynomial zero contour gives direct access to normals and curvature, so surface-tension and sharp-interface terms can be computed from the same reconstruction that conserves volume.
On unstructured triangular grids, the scheme reaches L1 errors lower than the compared geometrical VOF methods in the single-vortex test, suggesting that the algebraic tanh construction is not a compromise on unstructured meshes.
The semi-Lagrangian level set update removes the high-order WENO reconstruction used in earlier coupled THINC/level set schemes, which simplifies the algorithm while keeping sub-cell resolution.
The framework extends naturally to higher order interface polynomials, since the THINC scaling does not depend on a particular polynomial degree.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension the paper does not run is a three-dimensional case with topological change; the tanh scaling should in principle handle breakup and merger because the VOF cell average is conserved, but the reinitialization step would need to re-seed the level set after topology changes.
The conservation claim is exact only in the limit of large $\beta$; at finite $\beta$ the tanh function is a smeared Heaviside, so an editor would expect small volume drift on very long advection runs unless $\beta$ is increased with resolution.
The same two-way scaling could be used as a post-processing tool: given a VOF field from any method, invert the tanh to obtain a smooth level set field and extract curvature from the polynomial surface, an application the paper only hints at.
A stress test worth running is vortex deformation with the reverse flow turned off and the filament stretched beyond one cell width; the authors show such a case at $T/2$, but quantitative conservation of the under-resolved tail over long time is not measured.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things about this paper. First, the core idea is real: they use the THINC function as an exact analytic bridge between a VOF field and a level set field, and they exploit that bridge to update the level set semi-Lagrangianly without any WENO reconstruction. That is a concrete algorithmic simplification over their own coupled THINC/level set method (ref. 17), and it is the main reason to read the paper. Second, the paper does not actually demonstrate the thing its abstract promises most loudly—geometrical faithfulness. They report VOF L1 errors only, not normal, curvature, or distance-to-interface errors, and they never compare against the coupled THINC/level set method from the same group, which is the most relevant baseline. That is the soft spot, and it is real, but it is addressable rather than fatal.

What is genuinely new: the inverse THINC scaling formula (Eq. 9) is elementary, but using it to convert the cell-wise THINC reconstruction into a pointwise level set value at the departure point is a neat trick. It avoids the cost and complexity of high-order interpolation for the level set advection. The per-cell polynomial surface, PSI, is also a nice way to preserve sub-cell interface geometry while keeping the VOF conservation constraint exact. On the standard Zalesak and vortex benchmarks, the L1 errors are competitive with geometric VOF methods on both Cartesian and triangular grids. The derivation of the conversion formulas is clean, and the VOF constraint in Eq. (16) is a conservation constraint, not a fitted parameter. I have no circularity concern.

Where I would push back: the term "unify" is doing a lot of work. The scheme is a hybrid that alternates between VOF and level set updates, not a single evolution equation, but that is semantics and not a flaw. The more substantive worry is the one your stress-test flags. In interface cells, Step 4.2 produces level set values that are algebraic distances from the PSI, not signed distances, and Step 5 only reinitializes away from the interface. So the next polynomial fit in Step 1 mixes signed-distance and non-signed-distance values, which can bias the shape of the polynomial, not just its shift. For thin filaments below grid resolution, a single quadratic PSI per cell cannot represent two close interface branches, and the level set will likely smooth them. The paper's own Fig. 5 shows the tail is captured, but they never measure how faithfully. That is a genuine gap, and it weakens the central claim. I would also note the paper would be stronger with code or a reproducibility statement, and with a comparison to ref. 17.

All that said, the algorithmic idea is sound and the errors on benchmark advection tests are good. The paper deserves a serious peer review—a referee should push for quantitative geometry metrics (normal error, curvature error, reinitialization bias) and a head-to-head against the coupled THINC/level set method. If those come out well, this becomes a citeable contribution. So: send it to review, but insist on the geometry check.

Referee Report

4 major / 6 minor

Summary. This paper proposes a novel interface-capturing scheme, THINC-scaling, that unifies VOF (volume of fluid) and level set methods. The key idea is an analytic conversion between a level set function and a continuous Heaviside-like VOF function via the THINC function: the forward THINC scaling maps a level set polynomial to a VOF field, and the inverse THINC scaling maps the VOF field back to a level set field. The paper gives a solution procedure (Section 3) that reconstructs a cell-wise polynomial from the discrete level set, shifts it to satisfy the VOF conservation constraint, transports the VOF function by a finite volume scheme, updates the level set by a semi-Lagrangian step using the inverse THINC scaling, and reinitializes the level set away from the interface. Numerical tests are presented for Zalesak's solid body rotation and the Rider-Kothe single vortex on both Cartesian and triangular unstructured grids, with VOF L1 errors compared against several geometric VOF methods. The central claim is that the scheme simultaneously preserves the mass/volume conservation of VOF and the geometrical faithfulness of level set methods.

Significance. If the central claim is substantiated, the paper offers a practically useful unification: a conservative interface-capturing method with sub-cell geometric information that is simpler than PLIC-based VOF and avoids the conservation defect of pure level set methods. The analytic conversion formulas (6) and (9) are clean and exact, the VOF constraint (16) enforces conservation by construction, and the scheme extends naturally to unstructured grids. The benchmarks show VOF L1 errors competitive with or better than published geometric VOF methods. However, the geometric-fidelity claim is currently supported only by qualitative contour plots; no quantitative level set or geometric error is reported, and the treatment of the level set in interface cells raises a correctness question that the paper does not address. For these reasons the result is promising but not yet fully established.

major comments (4)

[Section 4, Eq. (26)] The quantitative assessment is limited to the VOF L1 error defined in Eq. (26). The abstract and Section 5 claim that THINC-scaling preserves the geometrical faithfulness of the level set method, but no error measure for the level set field, interface position, normal, or curvature is reported. Figures 4-6 show qualitative PSI plots only. Since geometric fidelity is the distinguishing advantage over algebraic VOF methods, please add quantitative geometry diagnostics, such as error in the zero-contour position relative to the exact solution, normal and curvature errors, or a measure of signed-distance accuracy in interface cells, for the Zalesak and vortex tests.
[Section 3, Steps 4.2 and 5] In an interface cell, the inverse THINC scaling (23) yields phi(x) = psi_id(x) = P_id(x) + phi_delta, which is an algebraic distance to the PSI, not a signed distance. Step 5 reinitializes only cells away from the interface, so these non-signed-distance values persist in interface cells. The next Step 1 then fits P_i to a stencil that mixes reinitialized signed-distance values with non-reinitialized algebraic-distance values; this mismatch can bias the polynomial fit, and Step 2 only shifts the polynomial rather than correcting its shape. The paper does not quantify this effect. Please verify that the level set remains a consistent signed-distance-like function in interface cells, for example by reporting |grad phi| in interface cells or by testing sensitivity to the reinitialization band.
[Section 3, Step 1 and Remark 6] The sub-cell geometric fidelity claim relies on a single polynomial P_i per cell. For a thin filament whose two interface branches pass through the same cell, as in the vortex test at t=T/2 (Fig. 5), a single polynomial zero set can represent the two branches only in special cases; in general the PSI will place the interface on an averaged curve. The manuscript provides no geometry error data for the thin tail region, so this limitation is not assessed. Please add a quantitative study of the filament region or explicitly discuss the topological limitation of a single polynomial representation.
[Section 4, Tables 1 and 2] No comparison is made with the coupled THINC/level set method of Ref. [17], although Remark 5 identifies the semi-Lagrangian inverse-THINC level set update as the essential difference from [17]. A direct comparison with [17] on the same benchmarks is needed to substantiate that the new update rule is at least as accurate as the HJ-WENO update it replaces.

minor comments (6)

[Abstract and Section 5] There are typographical errors: 'samectime' in the abstract and 'unstructured and unstructured grids' in Section 5 should read 'structured and unstructured grids'.
[Section 2 and Fig. 1] The caption of Fig. 1 contains garbled text ('0) x (/g32/g73'), and the phrase 'VOF velue' in Section 2 should be 'VOF value'.
[Section 2, Eq. (8)] The notation in Eq. (8), using dx, dy, dz and D, is unclear; a multi-index notation would make the constraint condition easier to read.
[Section 3, Step 2] Existence and uniqueness of the solution phi_delta to Eq. (16) are not discussed. Since the Newton iteration is referenced to [16], a brief remark on the monotonicity of the left-hand side in phi_delta would help.
[Section 4, Table 1] The header of Table 1 is corrupted ('Methods 642 Order 128 2 Order 256 2'); the formatting should be cleaned up.
[Section 4] The polynomial order p is fixed at 2 in all numerical tests, and no study of the dependence on p is reported. Since the paper emphasizes high-order interface representation, a p-convergence test or a discussion of the practical range of p would strengthen the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the THINC-scaling scheme is self-consistent and validated against external benchmarks.

full rationale

THINC-scaling is a numerical scheme, not an empirical model, and its derivation chain is self-consistent without reusing its inputs as predictions. The conversion formulas (6) and (9) are analytic bijections between the level set and THINC/VOF fields; the polynomial P_i in Step 1 is reconstructed from the discrete level set field, and the cell-wise THINC function (17) is constrained by the VOF value through Eq. (16), which is a conservation constraint rather than a fitted parameter. The level set update in Step 4.2 uses a semi-Lagrangian evaluation of the reconstructed field at the departure point; this is exactly the inverse-THINC evaluation of the same reconstruction, so it does not introduce an independent prediction that could reduce to the input by construction. The only free parameter, beta, is set by the usual THINC practice (Remark 7) and not tuned to benchmark results. The scheme is validated against external benchmarks (Zalesak [22], Rider-Kothe [4]) and compared with independent methods [23,24,25]. Self-citations to [13-17] provide prior implementation details and THINC background, but the central claim—mass-conservative and geometrically faithful interface capturing—does not rest on an unverified self-citation; the numerical tests and analytic formulas carry the argument. Therefore no circular step is identifiable.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The method does not introduce new physical entities. Its additional assumptions are standard in interface capturing: a signed distance field, polynomial surface reconstruction, and a tuned interface steepness. The main free parameter is beta, which controls the interface thickness and is set by hand; the volume conservation constraint (Eq. 16) is a well-posed projection rather than a fitted parameter.

free parameters (3)

Steepness parameter beta = 6 (structured), 6/Delta (unstructured)
Controls interface thickness; chosen based on the desired jump width (Eq. 25), not fitted to benchmark results. It affects the reconstructed interface profile and the inverse level set values.
Volume fraction threshold epsilon = 1e-8
Used to identify interface cells and define the range of the transition layer; conventional small value, does not affect the central derivation.
Polynomial order p = 2 (quadratic)
Interface is represented by a quadratic polynomial in the tests; higher orders are possible but not tested here, so the claimed high-order capability is partly unverified.

assumptions (3)

domain assumption The interface is advected by a prescribed velocity field without mass transfer, so the VOF and level set evolution equations (11) and (12) hold.
This is the standard advection model for interface capturing; it excludes phase change, surface tension effects in the transport, and topological changes beyond what the numerical diffusion creates.
domain assumption The level set function can be accurately approximated cell-wise by a polynomial of degree p with coefficients obtained from neighboring level set values via Lagrange interpolation or least squares.
The fidelity of the scheme depends on this polynomial fit. The paper does not analyze the approximation error, and under-resolved thin filaments (Fig. 5) are a stress test for this assumption.
ad hoc to paper The steepness parameter beta=6 gives a sufficiently sharp transition to approximate the Heaviside function while remaining differentiable.
This value is chosen based on experience with the THINC method (ref. 14) and is not derived from first principles; it is a tunable parameter that affects all results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of THINC-scaling scheme that unifies VOF and level set methods." pith.science (2026). https://pith.science/paper/Q2GXA773

@misc{pith2026190804529,
  author       = {Pith},
  title        = {Pith review of: THINC-scaling scheme that unifies VOF and level set methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q2GXA773}},
  note         = {Machine review of arXiv:1908.04529}
}

read the original abstract

We present a novel interface-capturing scheme, THINC-scaling, to unify the VOF (volume of fluid) and the level set methods, which have been developed as two completely different approaches widely used in various applications. The THINC-scaling scheme preserves at the samectime the advantages of both VOF and level set methods, i.e. the mass/volume conservation of the VOF method and the geometrical faithfulness of the level set method. THINC-scaling scheme allows to represent interface with high-order polynomials, and has algorithmic simplicity which eases its implementation in unstructured grids.

Figures

Figures reproduced from arXiv: 1908.04529 by the authors.

**Figure 2.** VOF 0.5 contour line in Zalesak solid body rotation [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** The numerical results of Zalesak solid body rotati [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Numerical results for Rider-Kothe single vortex t [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: The distorted interface at T/2 on 128 × 128 grid (a) where the boxed part is enlarged as the close-up of the thin tail in panel (b) which shows that the film of tail is under the resolution of grid cell, but can be still resolved by the THINC-scaling scheme. We tested …

**Figure 6.** Figure 6: Numerical results for Rider-Kothe single vortex t [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

[17]

L. Qian, Y . Wei, F. Xiao, Coupled THINC and level set meth od: A conservative interface capturing scheme with high-or der surface represen- tations, Journal of Computational Physics 373 (2018) 284–3 03

work page 2018
[1]

C. W. Hirt, B. D. Nichols, V olume of ﬂuid (VOF) method for t he dynamics of free boundaries, Journal of computational ph ysics 39 (1) (1981) 201–225

work page 1981
[2]

D. L. Y oungs, Time-dependent multi-material ﬂow with la rge ﬂuid distortion, Numerical methods for ﬂuid dynamics

work page
[3]

Lafaurie, C

B. Lafaurie, C. Nardone, R. Scardovelli, S. Zaleski, G. Z anetti, Modelling merging and fragmentation in multiphase ﬂows with SURFER, Journal of Computational Physics 113 (1) (1994) 134–147

work page 1994
[4]

W. J. Rider, D. B. Kothe, Reconstructing volume tracking , Journal of computational physics 141 (2) (1998) 112–152

work page 1998
[5]

Scardovelli, S

R. Scardovelli, S. Zaleski, Analytical relations conne cting linear interfaces and volume fractions in rectangula r grids, Journal of Computa- tional Physics 164 (1) (2000) 228–237

work page 2000
[6]

Osher, J

S. Osher, J. A. Sethian, Fronts propagating with curvatu re-dependent speed: algorithms based on Hamilton-Jacobi f ormulations, Journal of computational physics 79 (1) (1988) 12–49

work page 1988
[7]

J. A. Sethian, Level set methods and fast marching method s: evolving interfaces in computational geometry, ﬂuid mec hanics, computer vision, and materials science, vol. 3, Cambridge universit y press, 1999

work page 1999

Show all 25 references

[8]

Osher, R

S. Osher, R. Fedkiw, Implicit Functions, in: Level Set Me thods and Dynamic Implicit Surfaces, Springer, 3–16, 2003

2003
[9]

Sussman, E

M. Sussman, E. G. Puckett, A coupled level set and volume- of-ﬂuid method for computing 3D and axisymmetric incompres sible two-phase ﬂows, Journal of computational physics 162 (2) (2000) 301–3 37

2000
[10]

M´ enard, S

T. M´ enard, S. Tanguy, A. Berlemont, Coupling level set/VOF/ghost ﬂuid methods: V alidation and application to 3D simula tion of the primary break-up of a liquid jet, International Journal of Multipha se Flow 33 (5) (2007) 510–524

2007
[11]

Y ang, A

X. Y ang, A. J. James, J. Lowengrub, X. Zheng, V . Cristini , An adaptive coupled level-set /volume-of-ﬂuid interface capturing method for unstructured triangular grids, Journal of Computational P hysics 217 (2) (2006) 364–394

2006
[12]

D. Sun, W. Tao, A coupled volume-of-ﬂuid and level set (V OSET) method for computing incompressible two-phase ﬂows, International Journal of Heat and Mass Transfer 53 (4) (2010) 645–655

2010
[13]

F. Xiao, Y . Honma, T. Kono, A simple algebraic interface capturing scheme using hyperbolic tangent function, Inter national Journal for Numerical Methods in Fluids 48 (9) (2005) 1023–1040

2005
[14]

F. Xiao, S. Ii, C. Chen, Revisit to the THINC scheme: a sim ple algebraic VOF algorithm, Journal of Computational Phys ics 230 (19) (2011) 7086–7092

2011
[15]

S. Ii, K. Sugiyama, S. Takeuchi, S. Takagi, Y . Matsumoto , F. Xiao, An interface capturing method with a continuous fu nction: The THINC method with multi-dimensional reconstruction, Journal of Computational Physics 231 (5) (2012) 2328–2358

2012
[16]

B. Xie, F. Xiao, Toward e ﬃcient and accurate interface capturing on arbitrary hybrid unstructured grids: The THINC method with quadratic surface representation and Gaussian quadrature, Journal o f Computational Physics 349 (2017) 415–440

2017
[18]

Shu, Total-variation-diminishing time discret izations, SIAM J

C.-W. Shu, Total-variation-diminishing time discret izations, SIAM J. Sci. Stat. Comput. 9 (1988) 1073–1084

1988
[19]

Zhao, A fast sweeping method for eikonal equations, M athematics of computation 74 (250) (2005) 603–627

H. Zhao, A fast sweeping method for eikonal equations, M athematics of computation 74 (250) (2005) 603–627

2005
[20]

Dianat, M

M. Dianat, M. Skarysz, A. Garmory, A Coupled Level Set an d V olume of Fluid method for automotive exterior water manag ement applica- tions, International Journal of Multiphase Flow 91 (2017) 1 9–38

2017
[21]

Strain, Semi-Lagrange methods for level set equatio ns, Journal of Computational Physics 2 (1999) 498–533

J. Strain, Semi-Lagrange methods for level set equatio ns, Journal of Computational Physics 2 (1999) 498–533

1999
[22]

S. T. Zalesak, Fully multidimensional ﬂux-corrected t ransport algorithms for ﬂuids, Journal of computational ph ysics 31 (3) (1979) 335–362

1979
[23]

Maric, H

T. Maric, H. Marschall, D. Bothe, An enhanced un-split f ace-vertex ﬂux-based V oF method, Journal of computational physics 371 (2018) 967–993

2018
[24]

Owkes, O

M. Owkes, O. Desjardins, A computational framework for conservative, three-dimensional, unsplit, geometric tra nsport with application to the volume-of-ﬂuid (VOF) method, Journal of computational physics 270 (2014) 587–612

2014
[25]

Scheuﬂer, J

H. Scheuﬂer, J. Roenby, Accurate and e ﬃcient surface reconstruction from volume fraction data on g eneral meshes, Journal of computational physics 383 (2018) 1–23. 11

2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [17]

L. Qian, Y . Wei, F. Xiao, Coupled THINC and level set meth od: A conservative interface capturing scheme with high-or der surface represen- tations, Journal of Computational Physics 373 (2018) 284–3 03

work page 2018

[2] [1]

C. W. Hirt, B. D. Nichols, V olume of ﬂuid (VOF) method for t he dynamics of free boundaries, Journal of computational ph ysics 39 (1) (1981) 201–225

work page 1981

[3] [2]

D. L. Y oungs, Time-dependent multi-material ﬂow with la rge ﬂuid distortion, Numerical methods for ﬂuid dynamics

work page

[4] [3]

Lafaurie, C

B. Lafaurie, C. Nardone, R. Scardovelli, S. Zaleski, G. Z anetti, Modelling merging and fragmentation in multiphase ﬂows with SURFER, Journal of Computational Physics 113 (1) (1994) 134–147

work page 1994

[5] [4]

W. J. Rider, D. B. Kothe, Reconstructing volume tracking , Journal of computational physics 141 (2) (1998) 112–152

work page 1998

[6] [5]

Scardovelli, S

R. Scardovelli, S. Zaleski, Analytical relations conne cting linear interfaces and volume fractions in rectangula r grids, Journal of Computa- tional Physics 164 (1) (2000) 228–237

work page 2000

[7] [6]

Osher, J

S. Osher, J. A. Sethian, Fronts propagating with curvatu re-dependent speed: algorithms based on Hamilton-Jacobi f ormulations, Journal of computational physics 79 (1) (1988) 12–49

work page 1988

[8] [7]

J. A. Sethian, Level set methods and fast marching method s: evolving interfaces in computational geometry, ﬂuid mec hanics, computer vision, and materials science, vol. 3, Cambridge universit y press, 1999

work page 1999